Find the AI tells in your draft.

How to read the results

Each finding has a level. Strip means the line never belongs in a finished text: chat leftovers, placeholders, citation residue. Rewrite means fixing every instance, from the facts. Cap means the device is fine once in a while and a problem when it repeats, as dashes and fragments are. Watch means it only matters in clusters.

How often a pattern appears matters more than any single hit: one contrast or one dash in a 2,000-word essay proves nothing, and three in a 150-word email are what readers notice. Fix each one from the facts. A synonym for a buzzword changes nothing a reader feels.

What it isn’t

It isn’t a detector. It can’t tell who wrote a text and doesn’t try, because people use every one of these devices and plenty of AI text avoids them. The detectors that do claim to tell can be badly wrong: in a 2023 study, seven of them flagged 61% of essays by non-native English speakers as AI-written, and almost none of the essays by US eighth-graders. If you teach, use it to show students how to edit, never as evidence against one.

It isn’t a humanizer either. It changes nothing in your text, and it won’t help anyone hide who wrote something. It points at lines, and you decide what to do with them.

Why I made it

I write with AI. Before an essay goes up on this site, I run it through this checklist and rewrite what it flags, from the facts rather than with synonyms. The checker is that same list: the linter from my copywriter skill, which is open source, ported to run in your browser. The skill does the rest of the editor’s job for Claude Code and other agents. It drafts from facts, edits in separate passes and checks its own rewrites for new habits.

Every pattern, with examples

Signs of AI writing explains each pattern the checker looks for, and the ones it can’t see, with a before and after for each.